What Autodesk documents about Arnold GPU, how to decide between Arnold CPU and GPU for a farm job, and a Maya scene preparation checklist.
Maya Arnold cloud rendering can run on the CPU or the GPU, and the right choice for a cloud or farm job depends on the scene, the features it uses and the hardware you can rent. This guide explains what Autodesk documents about Arnold GPU, how to decide between CPU and GPU, and what to check before sending a Maya sequence to a farm.

Contents
Short answer
- Arnold GPU is built on NVIDIA’s OptiX framework and is optimized for NVIDIA RTX hardware, so it needs a compatible NVIDIA GPU. Arnold CPU has no such requirement.
- Autodesk states that you can switch between CPU and GPU rendering with the same settings, which makes it practical to test a frame both ways.
- GPU support covers OSL, OpenVDB volumes, on-demand texture loading, most light path expressions (LPEs), lights, shaders and all cameras. That is not the same as full parity with every CPU feature, so test your own shaders and AOVs on a short frame range before committing.
Arnold CPU or GPU: how to decide
Treat this as a measurement problem, not a rule. Render the same representative frame on both modes and compare time per frame, memory use and visual differences. Points that usually decide the outcome:
- Scene memory. GPU rendering is limited by VRAM, while CPU rendering uses system RAM, which is usually larger on server machines. Very heavy geometry, hair or large texture sets can favor CPU.
- Feature use. If a shader, AOV or LPE setup you rely on is not available on the GPU, you need CPU for the final render.
- Look development speed. GPU can be attractive for interactive work and quick test renders, then CPU or GPU for finals depending on the comparison.
- Cost per frame. Compare hourly price and time per frame, not hourly price alone. The render farm cost calculator shows the arithmetic.
Autodesk also notes that GPU noise is on par with CPU noise when adaptive sampling is used, which supports comparing both modes with the same sampling settings. The source for these statements is the 2019 Arnold 6 announcement coverage linked below; check the current Arnold documentation for the latest GPU requirements and supported features before you decide.
Maya Arnold cloud rendering: preparing a scene for a farm
- Use project-relative paths and collect every texture, reference and cache. Maya references that point to local drives are a common cause of missing-file errors on a remote node.
- Bake or export simulations that cannot be recomputed deterministically on the node, such as nCloth, nParticles and XGen caches, and confirm they load correctly from the shared location.
- Match the Maya, Arnold plugin and third-party plugin versions to the farm’s versions.
- Use the render layer or render setup you intend to ship, and confirm output naming, padding and format.
- Test a single frame and a short range on the target hardware before queuing the whole shot.
- Plan for failures. Long jobs lose frames, and a routine for finding and re-rendering them saves time; see how to find and re-render broken frames.
Studio and team workflows
For multi-artist teams the practical issues are asset versioning, a shared output location and consistent plugin versions across machines. A render manager helps with queues and priorities. Autodesk Arnold is available as a plug-in for Maya, 3ds Max, Houdini, Cinema 4D and Katana, so a team that mixes applications can often keep one renderer across them.
In practice, Maya Arnold cloud rendering comes down to three decisions. First, test one representative frame on both CPU and GPU before committing a sequence. Second, confirm that every feature the scene uses is supported on the device you plan to rent. Third, package textures, caches and plugins so the render node sees the same files as your workstation. Doing this on a short test job is the safest way to start Maya Arnold cloud rendering, on a managed farm or a rented workstation alike.
Related reading
- Why your animation render keeps crashing: troubleshooting guide
- GPU and VRAM for animation rendering
- How to choose the best render farm for your projects
- IaaS, SaaS and PaaS render farms explained
FAQ
Does Arnold GPU work on any graphics card?
Arnold GPU is built on NVIDIA OptiX and optimized for NVIDIA RTX hardware, so it needs a compatible NVIDIA GPU. Arnold CPU has no such requirement. Check the current Arnold documentation for supported cards.
Can I switch between Arnold CPU and GPU with the same settings?
Autodesk states you can switch between CPU and GPU rendering with one click while keeping the same settings, which makes it practical to test a frame both ways.
Does Arnold GPU support every CPU feature?
No. Coverage includes OSL, OpenVDB volumes, on-demand texture loading, most LPEs, lights, shaders and all cameras, but that is not full parity. Test your own shaders and AOVs on a short range first.
What should I check before sending a Maya scene to a farm?
Use project-relative paths, collect textures, references and caches, match Maya and plugin versions to the farm, confirm output settings, and render a test frame and short range on the target hardware.
Sources
- TV Technology: Autodesk Arnold 6 offers production rendering on both CPU and GPU (December 2019 announcement coverage)
COMMENTS